def FetchBU3DData(DataPath,
facePointMult3=21081,
fileCnt=1000,
printTime=False,
valCnt=400,
trainDirName="train_Resampled",
testDirName="val_Resampled",
landmarkDirName="landmarks"):
startTime = time.time()
if printTime:
print("Loading Data...")
X = np.zeros((fileCnt, facePointMult3 + 83 * 3))
Y = np.zeros((valCnt, facePointMult3 + 83 * 3))
Y_ = np.zeros((valCnt, facePointMult3 + 83 * 3))
cnt = 0
trainDataPath = os.path.join(DataPath, trainDirName)
testDataPath = os.path.join(DataPath, testDirName)
landmarkPath = os.path.join(DataPath, landmarkDirName)
# print(landmarkPath)
list_dirs = os.walk(trainDataPath)
for _, _, files in list_dirs:
for f in files:
if f[-3:] == 'xyz':
if cnt >= fileCnt: break
bndFile = f[:-3] + 'bnd'
xyzData = dataio.loadData(os.path.join(trainDataPath, f))
bndData = dataio.loadData(os.path.join(landmarkPath, bndFile))
F = combineData(xyzData, bndData)
X[cnt, :] = F.ravel()
cnt += 1
list_dirs = os.walk(testDataPath)
cnt = 0
for _, _, files in list_dirs:
for f in files:
if f[-3:] == 'xyz':
if cnt >= valCnt: break
bndFile = f[:-3] + 'bnd'
xyzData = dataio.loadData(os.path.join(testDataPath, f))
bndData = np.zeros((83, 3))
bndTruth = dataio.loadData(os.path.join(landmarkPath, bndFile))
F = combineData(xyzData, bndData)
F_GT = combineData(xyzData, bndTruth)
Y_[cnt, :] = F.ravel()
Y[cnt, :] = F_GT.ravel()
cnt += 1
print("Data loaded.\nTrain Shape:", X.shape, " | Test Shape:", Y.shape)
if printTime:
print("Cost {} seconds.".format(time.time() - startTime))
return X, Y_, Y #Train, toLearn, GT
def demo(readLocation="../data/sensor_data/late_small.dat"):
import dataio
oData = dataio.loadData(readLocation)
val = bestCorrelation(oData.data, offsetMax=3, offsetMin=1)
neighbors = createNeighbors(val)
oData.correlation = val
oData.neighbors = neighbors
dataio.saveData(readLocation, oData)
def demo(readLocation = "../data/sensor_data/late_small.dat"):
import dataio
oData = dataio.loadData(readLocation)
val = bestCorrelation(oData.data, offsetMax = 3, offsetMin = 1)
neighbors = createNeighbors(val)
oData.correlation = val
oData.neighbors = neighbors
dataio.saveData(readLocation, oData)
def FetchAllData(TrainDataPath):
# ExportFace = "ExportBnd"
curPath = os.getcwd()
FileCnt = 1000
X = np.zeros((FileCnt, 23349))
list_dirs = os.walk(TrainDataPath)
i = 0
for root, _, files in list_dirs:
for f in files:
if f[-3:] == 'xyz':
if i >= FileCnt: break
# print(f)
bndFile = f[:-3] + 'bnd'
xyzData = dataio.loadData(os.path.join(TrainDataPath, f))
bndData = dataio.loadData(os.path.join(TrainDataPath, bndFile),
spliter='\t\t')
# print(bndData.shape)
F = combineData(xyzData, bndData)
X[i, :] = F.ravel()
i += 1
# print(X)
return X
def FetchXYZData(TestDataPath):
curPath = os.getcwd()
FileCnt = 20
X = np.zeros((FileCnt, 23349))
list_dirs = os.walk(TestDataPath)
i = 0
for root, _, files in list_dirs:
for f in files:
if f[-3:] == 'xyz':
if i >= FileCnt: break
print(f)
xyzData = dataio.loadData(os.path.join(TestDataPath, f))
bndData = np.zeros((83, 3))
#print(bndData.shape)
F = combineData(xyzData, bndData, normalize=False)
X[i, :] = F.ravel()
i += 1
# print(X)
return X
#directories.append("detectionsOct2007")
#directories.append("detectionsNov2007")
#directories.append("detectionsDec2007")
#directories.append("detectionsJan2008")
#directories.append("detectionsFeb2008")
#directories.append("detectionsMar2008")
#directories.append("detectionsApr2008")
directories.append("detectionsJan2008-Mar2008")
#directories.append("detectionsSep2007-Dec2007")
for i in range(1, 11):
for j in range(0, 5):
sensors.append(i*10 + j)
print "Sensors:" + str(len(sensors))
tdMatrix = calculateTDMatrix(readLoc, startTime, endTime, interval, directories, sensors)
print tdMatrix
oData = None
try:
oData = dataio.loadData(fileLoc)
except:
oData = bbdata.Dataset(None)
oData.tdMatrix = tdMatrix
dataio.saveData(fileLoc, oData)
import markov_anneal
import dataio
import warnings
import os
import random
import suppress
warnings.simplefilter("ignore")
readLocation = "../data/sensor_data/small_54_64.dat"
splitLen = 8
if __name__ == "__main__":
oData = dataio.loadData(readLocation)
obs = 2**oData.data.shape[1]
states = splitLen
sData = markov_anneal.splitActivityMax(oData.cd[0:50000], splitLen)
scores = [0, 0]
entropys = [0, 0]
for i in range(2, 26):
print "Models:" + str(i)
bestScore = -1
bestModels = []
bestData = []
bestOut = []
import markov_anneal
import dataio
import warnings
import os
import random
import suppress
warnings.simplefilter("ignore")
readLocation = "../data/sensor_data/small_54_64.dat"
splitLen = 8
if __name__ == "__main__":
oData = dataio.loadData(readLocation)
obs = 2**oData.data.shape[1]
states = splitLen
sData = markov_anneal.splitActivityMax(oData.cd[0:50000], splitLen)
scores = [0, 0]
entropys = [0, 0]
for i in range(2, 26):
print "Models:" + str(i)
bestScore = -1
bestModels = []
bestData = []
bestOut = []
#directories.append("detectionsOct2007")
#directories.append("detectionsNov2007")
#directories.append("detectionsDec2007")
#directories.append("detectionsJan2008")
#directories.append("detectionsFeb2008")
#directories.append("detectionsMar2008")
#directories.append("detectionsApr2008")
directories.append("detectionsJan2008-Mar2008")
#directories.append("detectionsSep2007-Dec2007")
for i in range(1, 11):
for j in range(0, 5):
sensors.append(i * 10 + j)
print "Sensors:" + str(len(sensors))
tdMatrix = calculateTDMatrix(readLoc, startTime, endTime, interval,
directories, sensors)
print tdMatrix
oData = None
try:
oData = dataio.loadData(fileLoc)
except:
oData = bbdata.Dataset(None)
oData.tdMatrix = tdMatrix
dataio.saveData(fileLoc, oData)
import myICP
import dataio
data1 = dataio.loadData("1_calib.asc")
data2 = dataio.loadData("2_calib.asc")
_, _, data2_ = myICP.icp(data2,
data1,
maxIteration=50,
tolerance=0.00001,
controlPoints=1000)
dataio.outputData("2_icp.asc", data2_)
st = "2008-02-01 00:00:00"
et = "2008-03-31 23:59:59"
st = datetime.datetime.strptime(st, "%Y-%m-%d %H:%M:%S")
et = datetime.datetime.strptime(et, "%Y-%m-%d %H:%M:%S")
sensors = [53, 52, 51, 50, 44]
validDays = [0, 1, 2, 3, 4, 5, 6]
compress = 2
counts = [0] * len(validDays)
splitLen = 8
tdMatrix = []
if __name__ == "__main__":
mData = dataio.loadData(modelFile)
mData.matrixToModel(mData.modelList)
models = mData.models
for d in validDays:
print "Day " + str(d)
print " Getting data."
#Iterate over all valid days.
cd, td = bbdata.comp(st, et, \
vDays = [d], \
comp = compress, \
sens = sensors)
print " Splitting."
#Get the split calculation finished.
sData = markov_anneal.splitActivityMax(cd, td, splitLen)
